Cool eval. Simply ask an LLM “Land or Water?” and give it a latitude and longitude coordinate as text. Ask 16,200 times, plot as image. The models know. From compressing the internet.
@Saccc_c@canghe@woniu20758393 Hey Sac, I’m building voice-controlled computer use on the Mac with Jev and Codex and came across Jev-cu. It seems very close to the area I’m working in. Would be good to compare notes sometime if you’re up for it.
@savboj Hey Sav, I’m building voice-controlled computer use on the Mac with Jev and came across Jev Use. Seems like we’re working around a lot of the same problems. Would be good to compare notes sometime if you’re up for it.
Id like to have a tiny Jev style model running locally as a safety later for my agents, it would check request and command against a set of rules i define. Stopping me from accidentally sharing personal data to the api or running unsafe commands
just asked codex to go over every one of my old chats where i discuss and plan travel and asked it to infer a list of my preferences based on this, it responded with a very accurate list. Now im adding this to memory so every future travel planner can filter by this list before showing me any suggestions
Voice mode from @openai is the single best learning tool out there.
1. Ask it to find sources on the topic
2. Get it to explain the concept
3. You explain back your understanding
4. It tells you where your understanding is off
5. Repeat steps 2-4